The long-term goal of this project is to understand how human listeners learn to better discriminate sounds. The particular objective of the proposed experiments is to test and refine a conceptual working model of learning on basic auditory tasks. The model predicts how learning to make a discrimination between two given sounds with simple temporal and spectral structures will influence performance on new sounds or new discrimination tasks. The model also predicts how different training schemes will increase or decrease this influence. For example, consistent with preliminary data, the model predicts that listeners who learn to discriminate the frequencies of two tones at one frequency, will behave as though they have not previously learned on that task when tested at a new frequency. To test the assumptions of the model, experiments are proposed to examine how learning to perform one discrimination task with one sound affects subsequent performance on that trained task with untrained sounds, and on untrained tasks with the trained sound. The information gained from this project can be expected to (1) supply baseline information for comparison with learning and generalization on tasks employing more complex sounds such as speech, (2) affect the design of traditional psychoacoustic experiments that do not focus on learning but are nevertheless influenced by it, (3) contribute to our understanding of the mechanisms underlying performance on particular trained tasks, (4) influence the development of therapeutic training schemes for individuals with auditory disorders, (5) inform the investigation of perceptual learning in other sensory systems, and (6) provide insights into the neurobiology of learning and memory.